Harvard alumni-founded AI search engine Exa completes $250 million Series C funding, valuation triples in a year

robot
Abstract generation in progress
AIMPACT News, May 21 (UTC+8), according to Beating Monitoring, Exa, which is building underlying search infrastructure specifically for AI agents, announced the completion of a $250 million Series C funding round led by a16z. The post-money valuation soared from $700 million in September last year to $2.2 billion. The funds will be used to expand proprietary index libraries, train a new generation of retrieval models, and scale the underlying throughput to hundreds of thousands of concurrent requests per second. Exa emphasizes its core moat as "full-stack self-built": possessing an independent crawler, a native index system tracking over 500 billion URLs, self-developed embedding models, and the third-generation vector database, completely moving away from the industry’s common "Google clone" approach. This AI-designed engine supports response times from 200 milliseconds (Exa Instant) to minutes-level autonomous deep retrieval (Deep Max, recently achieving SOTA in DeepSearchQA). Recent research also shows that replacing traditional search engines with Exa for RL agent training can save 69% of Token consumption, transforming search from a "reasoning tool" into an indispensable "training foundation." Currently, Exa’s monthly query volume has reached 1 billion, a tenfold increase in a year, serving top AI companies including Cursor and Cognition. Ironically, Exa, which once erected a billboard claiming "10^18 > 10^100" (Exa > Google), has recently become one of the first partners to integrate with Google Cloud Vertex AI, providing Grounding search deployment services for Gemini enterprise clients. (Source: BlockBeats)
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • 9
  • 12
  • Share
Comment
Add a comment
Add a comment
GovernanceVotingTug-Of-WarKing
· 1h ago
All vector databases are self-developed, tough person
View OriginalReply0
GlassDome
· 6h ago
2.2 billion valuation, backed by a16z, this narrative is fully hype.
View OriginalReply0
GateUser-1c5ab2b5
· 6h ago
Deep Max self-service deep search sounds like giving AI a microscope.
View OriginalReply0
BluePeonyDoesn'tDrop
· 7h ago
Using Cursor all the time, that does have some substance.
View OriginalReply0
BridgeTroll
· 7h ago
Becoming Gemini's Grounding, this move is brilliant.
View OriginalReply0
GateUser-34d2b0ab
· 7h ago
From Instant to Max, the product line is divided quite finely.
View OriginalReply0
NekoOnCall
· 7h ago
RL Province 69% Token, saving money is earning money.
View OriginalReply0
TheClarityAfterLiquidating
· 7h ago
5 trillion URLs, just thinking about the crawling cost is terrifying
View OriginalReply0
GateUser-0b71fc11
· 7h ago
Building your own index database is the true barrier; a shell company will eventually be exposed.
View OriginalReply0
View More
  • Pinned